d1-ultra

d1-ultra accepts text and images through Liquid's decisions API.

What the release provides

The larger hosted model uses an open-weight backbone with Liquid-specific modifications; the documentation does not identify that backbone. Charges use input tokens, and each question includes its supplied images.Liquid's hosted d1 decision models

How to use it well

Practical guidance. Use the same outcome definitions when comparing hosted models. An option probability needs task-specific calibration before it controls an automated action.

Specifications & access

d1-ultra specifications
DetailValue and conditions
AccessAPI
Model identifiersd1-ultra
Context window32,768 tokens

Hosted decision API

Liquid API model catalog